A comparison was made between the procedure of factoring variables before using them in discriminant analysis and the usual procedure of using the original variables in discriminant analysis. The results indicated factoring seven scale measurements on sockeye salmon (Oncorhynchus nerka) with varimax rotation produced four new variables which gave more valid results in classifying sockeye salmon by area of origin than the original variables when discriminant functions were computed. Although the results on the basic data from which the functions were derived were not as good using the four factor scores as the seven original variables, the accuracy of classification was much more consistent in the test data with the factor scores. The loss in accuracy was at leasttimes as great for functions based on the original variables as for ones based on factor scores. The errors in classifying fish to their individual places of origin were perhaps too large for the procedure to be useful in the field, but the accuracy of classification to the British Columbia or Alaska region was quite high.